Meeting update

This past week I have:

*Read the Hannebelle paper (very good/pertinent to what I am doing now).

*Added Poisson BC parameters under the tutorials>data files explained>global.data

And worked on the following sims:

*Reflecting quandrant of sphere:

I learned more about the role of BCs in fluid simulations. That is, how they control the fluid variables across boundaries. For instance, when slicing through a symmetric object, the boundaries on the faces of the slice should be reflecting so the flow at that boundary is the same, as it would be by symmetry. I also learned more about why using extrapolating BCs at a boundary for subsonic flow may lead to issues — it has to do with the ghost region copying data from adjacent domain cells. If the fluid variables copied are pointing in the 'wrong' direction, which may happen from random thermal motions in a subsonic flow, they will be pasted inside the neighboring domain cell and subsequently the flow will be flowing in the wrong direction inside this boundary. I learned that the winds module when used will set any inflow to be 0, but will allow outflow. With these in mind, I set the domain BC's to be reflecting on the 3 faces of the sliced sphere, and used the winds module on the other sides (periodic was NOT an option, as you need a pair of boundaries for periodic to work!). As far as the elliptic boundaries go, I am using reflecting as well on the same 3 sides, and multipole everywhere else. I am unsure of how to proceed from here, given that sim. is currently stuck (see ticket - ​https://clover.pas.rochester.edu/trac/astrobear/ticket/228)

*Larger box size, higher refinement inside the BE sphere:

I learned why refinement was not being triggered in my matched ambient simulation (see ticket 223). After making an adjustment to my problem module, am now running this set up on the larger domain (30.83) with 6 levels of refinement. This is refining the clump to about 33 cells per radius. I am running the sim out to t=0.2, which is about 2x longer than the time it took for Pthermal>Pcrit, and with 200 frames. This will give us a temporal resolution ~ 10 x greater than before so we can see more clearly the cause of the anomalous "bounce" that concerned us before.

I am also rerunning the Light ambient case at the same resolution, 303 + 6 levels.

Both of these sims are using the winds module at the domain boundary, and multipole for the elliptic.

Concerning refinement of the matched case. If we are interesting in refining the clump down to 10% its original size with 30 cells p/ radius, 7 levels of refinement would be sufficient. However, this seems a HIGHLY inefficient way to refine as I am currently refining now — with a geometric condition. A much better way would be to use Jean's length again, but to figure out how to restrict the Jeans refinement to be only within the sphere (not the Jeans unstable ambient medium). Will talk to Jonathan about this.

The matched ambient case is more than half way done now (external pressure on the sphere's edge should exceed threshold by now), and the light is ~ 1/10 the way. Hope to post figures/movies with the results by the meeting tomorrow — at least what I'll have so far by then.

Comments

No comments.